As you know,this is really a big topic,and it also a hard question can be resolved.Handling it,need lots of experiences,the skill of designing test case and maybe well-informed for project.Furthermore, I can't illuminate it clearly yet.However,I just want to state my view.Certainly,you may have reserve point,please let me know.Thanks!
It's well-known that we always use servel specs to cover one function,as it contain different methods to finish it.For instance,to modify a value,we can change it directly when it is generated,or we also can correct it use other user.Furthermore,we can use dissimilar user to check if they are mended completely.Suppose,if we have two user can change the value,and have three user can check its value.That can write down 8 specs.If we have more users,how about the result?Definitly,the number of specs will increase by progressional.That's really terrible.It is predictable that we must take lots of time on one function.Sure,I'm not saying that we shouldn't pay time and energy on only one issue,but can we use less cases to cover it.As you know,we also can execute less spec to discover issue.I think the important point of this question is the effictive coverage for that point.We should collect all users who can modify the value and gather other user who can check it.Then we should use an effectual method to cover that point.The theory is like the Orthogonal Matrix in math.I hold that it use crossing testing.But for now,I can't illustrate it clearly.We will settle this question in the future.
In short,we try to use less spec to cover being tested point effectively.I still study it,so I'll continue this article in the future.If you are insterested in this question,you need wait for a moment,I'm sorry for that.Additional,maybe you think this question is fit for mature system.If the project isn't steady,maybe this approach will miss bugs.But I think if we can use it correctly,it also can go through every possibility.For this point,I'll ask my friends to explain next time.Thanks your attention!
